Claude
Anthropic describes a model-improvement setting for consumer chats and says incognito chats are not used to improve Claude.
The cited current help page explains what happens when users allow chats or coding sessions to improve Claude, and points to privacy settings for control.
Adjust privacy and model-improvement settings in Claude privacy controls; incognito chats are excluded from model improvement.
Anthropic advises users to be thoughtful about highly sensitive information and describes privacy protections when chats are used for improvement.
The cited article explains protections but does not provide one short retention answer for ordinary users.
Anthropic says access is limited to a small number of personnel involved in model training when users allow improvement.
Foundational reference pages for teams adopting, deploying, and operating production AI with inspectable controls.